Daniel Lemire

This will be part of the next release.

The PR would be very welcomed. The simdjson library has an overall design that is similar to simdutf and many of the problems you discuss have been solved in simdtuf....

Work should continue in branch `riscv_develop`.

Seems reasonable to me but I am still preoccupied by the complexity. Unavoidably, a good chunk of the container code will need to be modified to accept new parameters. This...

@Dr-Emann You are a member of the core team. If you want to do this, I don't think anyone should have the authority to stop you. I trust that if...

Please run the following Python script: https://gist.github.com/lemire/fabe642b256987f94695c03b3d7c2515

Running my own script I do spot a significant regression on the IterateAll64 benchmark which is not part of your results. And a small degradation in RandomAccess64. I do not...

To be clear, my belief is that this affects IterateAll64 and RandomAccess64. I think that _IterateAll_ and _RandomAccess_ are **unaffected**. I trust that @SLieve is likely right that it is...

@madscientist We should understand why there is a performance degradation and what is actionable. IterateAll64: ``` 41.74% bench libroaring.so.4.0.0 [.] roaring64_iterator_advance ◆ 21.24% bench libroaring.so.4.0.0 [.] container_iterator_next ▒ 15.52% bench...

So I have opened an issue... https://github.com/RoaringBitmap/CRoaring/issues/755 There are three functions that I believe should be candidate for inlining. They are not currently. This might help with this issue.